Filter the signal waveform

[rahulmore]

Hi everyone.

I'm a beginner on Labview. I would like to ask about my problem described below:

I am analyzing a signal received from machine. It contains some kind of noise in the waveform and also some variations in amplitude along the waveform at the higher amplitude levels.
I will like to ask here about what will be best possible combination of filter to smooth-out this signal so that the it to have the same higher magnitude along it's path. I have marked the part in the attached file.
Thanks in advance for kind suggestions and solution.

- Rahul

[MaveenD]

hi Rahul,

You can get a clear idea if you try n workout with this filters

http://www.ni.com/white-paper/4851/en

[rahulmore]

The information on this white paper gives information about Labview full/professional and DFD toolkit, but still the use of correct filter combination is not clear from this information. Will you please elaborate?

- Rahul

[camerond]

It looks like you have some high frequency noise on top of your signal and the whole thing has an oscillation with a period of about 0.026 sec underneath it. Are you talking about minimizing both of these? Looks like to me that you'll need a bandpass filter, You'll have to experiment a bit with the type of filter and your passband to see how much of either you can eliminate/tolerate and how much distortion of your peaks you can accept.

Cameron

p.s. I like your units (-Rahul)

[johnsold]

Rahul,

That interfering signal is huge: 7 V on a signal of ~ 0.1 V. The first thing to do is to try to determine what causes it. (Correlate with rotational frequency of machine, for example?) Once you know that, then see if you can change the configuration of your transducers to eliminate or minimize the pickup of the signal.  It is always better to keep the interference out of the signal than to try to remove it later.  It also appears that the negative spikes occupy about 10-20% of the duration of the signal.  This means that you are losing that much of the information and it is unlikely that any signal processing will recover it.

The variations in amplitude along the waveform at the higher amplitude levels appear to be at 50 Hz, so power line frequency is the principal suspect here if you are in a part of the world which uses that frequency for the power mains. The issue there could be ground loops or inductive pickup. Again proper configuration of the transducers and signal conditioning is the preferred method of reducing this interference.

Tell us more about the equipment you are using and what you are trying to measure.

Lynn

[rahulmore]

Hello Cameron,
It seems from the signal that the frequency of this waveform nature is of 49-50 Hz. I am trying to work with bandpass filter but the result is still the same.
- Rahul

[rahulmore]

Hello Lynn,

The concept behind the design of this equipment is to measure the chips generated during the drilling process. It consists of a Light source and photo detector. The light is reflected at the groove of the tool and gives one negative spike in the signal for each reflection. And when it comes across a chip, the signal shows some kind of disturbance. Here my interest is in defining the standard signal and thus identifying the disturbances (Chips) in the signal.
You are right that the variations in amplitude along the waveform at the higher amplitude levels appear to be at 50 Hz. probably, it will be having some kind of relation with the Power line frequency. I will check that too.

But I will also like to define the filter parameters to remove this waveform nature and receive the signal so that it will help me to define standard signal pattern.
Thanks in advance if you have some suggestions for me.

Rahul

[camerond]

You say you have tried to work with a bandpass filter "but the result is still the same." Please attach your VI(s) to a post so we can see what you have tried - only the stuff connected to this problem. It would also be helpful if you could send a file (spreadsheet, text, whatever) with the time/signal data you show in your original post (or similar).

Are you saying that the short-term noise on top of the signal is not a problem? If so, you don't need the high frequency part of any filter you use.

But, then. looking at your reply to Lynn, why are you bothering with filtering out noise and background drift anyway? You signal is so huge, I don't see how you can have trouble detecting it with such a small background. If you don't know how to eliminate false-positive detection, just lower your sensitivity.

Cameron

[rahulmore]

Cameron,

I am interested to recover this waveform nature to get plain wave nature at higher amplitudes so that I can measure the disturbances (negative spikes) and define a standard profile of signal with neagtive spikes and deteremine if there are any changes in the pattern of negative sipkes. At the end, I am expecting the signal with nearly constant / almost constant higher amplitude level. If you have any suggestion, please share.

In the attachement, you will find
ChipSignal.tdms - The signal to be filtered out.
bandlimits.tdms - randomly identified band limit for the negative spike profile
filter.vi - trying to use filter combinations
Pattern Identification -XXXXX.tdms - trials done to separate out the major disturbances in negative spikes.

- Rahul

[rahulmore]

In some papers, I read that some kind of Adaptive noise cancelling is used in ECG to compensate for powerline frequency. It used kind of Steepest Descent Least Mean Square algorithm to modify filter weight. But I dont know how can i use it for my problem. Assistance requested.
- Rahul